Financing: the work was carried out under the grant from the Academy of Sciences of the Republic of Tatarstan, provided to young candidates of sciences (postdoctoral fellows) so that they could defend their doctoral dissertation, perform research and labor functions in scientific and educational organizations of the Republic of Tatarstan within the framework of the Scientific and Technological Development of the Republic of Tatarstan State Program.

Acknowledgements: the authors express their sincere gratitude to the Academy of Sciences of the Republic of Tatarstan for financial support of this study based on the results of a competitive selection (grant No. 153/2024-PD dated December 16, 2024) for ‘Ensuring the technological sovereignty of the healthcare system by criminal means’ research project, as well as deep gratitude to the staff of the Scientific Research Institute of Digital Technologies and Law of Kazan Innovative University named after Timiryasov VG for valuable consultations and constructive discussion of the conceptual provisions of the study.

Author contribution: Begishev IR — research conceptualization; development of predictable harm theoretical foundations; analysis of specific risks associated with using artificial intelligence in medical devices; systematization of predictable harm typology; research of legal framework for regulating AI systems in healthcare with various jurisdictions; formulation of research conclusions; preparation of the manuscript initial version; Shutova AA — development of a methodology for predicting and preventing predictable harm; creation of a matrix for assessing predictable harm at various stages of medical AI system life cycle; developing recommendations for practical implementation of predictable harm concept; analysis of literature sources; editing and critical revision of the manuscript with introduction of valuable intellectual content; visualization of research (making tables).

Compliance with ethical standards: a meeting of the ethics committee was not required, since this study is theoretical and methodological in nature and analyzes open literature sources and regulatory legal documents, without conducting experiments involving humans or animals and without using personal patient data.